Production modeling: PVWatts defaults vs a detailed panel and inverter model — why numbers differ
Two production estimates for the same roof can differ by several percent without either being wrong. In our engine, PVWatts with its default settings gave 1,638 kWh a year per kW of panels facing south at 20° in West Covina; the same four systems modeled with each panel's and inverter's own parameters gave 1,697 to 1,769 kWh per kW, 3.1% to 8.0% more across the two sites. The largest part of the gap is one setting: PVWatts assumes 14% system losses, the detailed model 10.5%, which alone adds 4.0%. The panel model adds another 0.8% to 2.7%, and the inverter model moves the result by -1.7% to 1.3%.
In this guide
Three models, four systems#
Swipe to see all columns →
| Site | System | PVWatts defaults, kWh per kW | Panel model, kWh per kW | Panel and inverter model, kWh per kW | Panel model effect | Inverter model effect | Total vs PVWatts |
|---|---|---|---|---|---|---|---|
| West Covina | Q.PEAK DUO 410 on IQ8HC | 1,638.3 | 1,717.2 | 1,737.4 | +0.8% | +1.2% | +6.1% |
| West Covina | REC Pure-RX 460 on IQ8X | 1,638.3 | 1,750.5 | 1,768.8 | +2.7% | +1.0% | +8.0% |
| West Covina | LONGi 400 on IQ8PLUS | 1,638.3 | 1,717.7 | 1,696.9 | +0.8% | −1.2% | +3.6% |
| West Covina | JA Solar 405 on Powerwall 3 | 1,638.3 | 1,719.2 | 1,707.1 | +0.9% | −0.7% | +4.2% |
| Lancaster | Q.PEAK DUO 410 on IQ8HC | 1,861.0 | 1,951.0 | 1,976.8 | +0.8% | +1.3% | +6.2% |
| Lancaster | REC Pure-RX 460 on IQ8X | 1,861.0 | 1,985.3 | 2,007.1 | +2.6% | +1.1% | +7.9% |
| Lancaster | LONGi 400 on IQ8PLUS | 1,861.0 | 1,950.6 | 1,918.3 | +0.8% | −1.7% | +3.1% |
| Lancaster | JA Solar 405 on Powerwall 3 | 1,861.0 | 1,952.4 | 1,953.0 | +0.9% | +0.0% | +4.9% |
Output per kW of panel rating for 12 panels facing south at 20°, typical-year weather at ZIP 91790 (West Covina) and ZIP 93535 (Lancaster). PVWatts defaults: a standard panel with a -0.37%/°C coefficient, 14% losses, a 96% inverter at DC/AC 1.2. Panel model: the panel's CEC single-diode parameters with the engine's 10.5% losses and the PVWatts inverter. Panel and inverter model: plus the inverter's CEC Sandia model. Panel model effect excludes the loss setting. Source: Cali-Graf engine.
Where the difference comes from#
- The loss setting. PVWatts' 14% default bundles soiling, shading, mismatch, wiring, connections, light-induced degradation, nameplate rating and availability. The engine uses 7.8% for the items that apply to its microinverter-based designs, plus a 3% shading allowance when no shading analysis is run. That choice alone moves the estimate by 4.0%, more than anything else here. - The panel. A specific panel's CEC parameters replace PVWatts' standard panel: its temperature coefficient, its behavior in low light and its rating. Panels with smaller temperature losses, such as the REC Pure-RX, gain the most. - The inverter. A microinverter that runs above PVWatts' flat 96% efficiency and rarely clips adds about a percent; one that clips on bright hours, such as an IQ8PLUS under a 400 W panel, takes about a percent or more away. A large hybrid inverter running at low load loses a little of its efficiency.
What to check when two estimates differ#
- Losses first. Compare the total loss percentage; a few points of losses explain most differences between tools. - Then the equipment. A tool that models the actual panel and inverter can differ from a generic model by a few percent either way. - Then the weather and shading. Different weather years and shading methods move results again, and are not compared here.
Data limits#
Four systems at two sites, all unshaded, facing south at 20°. The engine's loss setting is a modeling choice, not a measured value for any one home. The panel model effect is computed after removing the loss-setting effect by its ratio. Real systems also lose output to aging, which none of these single-year estimates include.
Why is my PVWatts estimate lower than my installer's?
In our runs the loss setting was the largest single difference: PVWatts assumes 14% losses by default; a model with 10.5% gives 4.0% more before any equipment difference. Check the loss percentage in both estimates first.
How much does the panel model matter?
In our four systems, 0.8% to 2.7% once the loss setting is set aside.
Is PVWatts accurate enough for a proposal?
It is a consistent, published method; its default losses make it conservative compared with a detailed model of the actual equipment, by 3.1% to 8.0% in our runs.
Methodology#
The Cali-Graf engine runs PySAM's PVWatts v8 with its defaults (standard module, 14% losses, 96% inverter efficiency, DC/AC 1.2, roof mount) and a detailed model in which each panel is the CEC single-diode model from its CEC parameters, the cell temperature comes from PVWatts' roof-mount model, and the inverter is the Sandia model with CEC parameters, per panel for microinverters. NSRDB typical-year weather at each ZIP. The loss-setting effect is (1 − 10.5%) ÷ (1 − 14%) − 1. Rounded half up.
Sources#
- NLR — PVWatts Calculator (Version 8). https://pvwatts.nlr.gov/
- NLR — PySAM, the Python package for the System Advisor Model. https://nrel-pysam.readthedocs.io/
- California Energy Commission — Solar Equipment Lists (PV modules and grid support inverters). https://solarequipment.energy.ca.gov/
- Sandia National Laboratories — Performance Model for Grid-Connected Photovoltaic Inverters (SAND2007-5036). https://www.osti.gov/biblio/920449
- NLR — National Solar Radiation Database (NSRDB), typical meteorological year. https://nsrdb.nlr.gov/
How to cite this page#
Cali-Graf. “Production modeling: PVWatts defaults vs a detailed panel and inverter model — why numbers differ.” cali-graf.com, updated Nov 18, 2026. https://cali-graf.com/guides/production-modeling-pvwatts-vs-pysam-vs-vendor-tools-why/